Stalled tasks with no remaining retries need a clean up mechanism

[maxcountryman]

If a task is stalled (i.e. outside of the heartbeat interval) but does not have remaining retries it should not be dequeued and therefore will be in an unrecoverable state.

While this is hopefully an uncommon state, it still needs to be addressed by some kind of cleanup mechanism. A periodic task which looks for such rows and marks them as failed should suffice.